Changes in circulating microRNAs-99/100 and reductions of visceral and ectopic fat depots in response to lifestyle interventions: the CENTRAL trial

ABSTRACT

Background

MicroRNAs (miRNAs) are short noncoding RNAs and important posttranscriptional regulators of gene expression. Adipose tissue is a major source of circulating miRNAs; adipose-related circulating miRNAs may regulate body fat distribution and glucose metabolism.

Objectives

We investigated how changes in adipose-related circulating microRNAs-99/100 (miR-99/100) in response to lifestyle interventions were associated with improved body fat distribution and reductions of diabetogenic ectopic fat depots among adults with abdominal obesity.

Methods

This study included adults with abdominal obesity from an 18-mo diet and physical activity intervention trial. Circulating miR-99a-5p, miR-99b-5p, and miR-100-5p were measured at baseline and 18 mo; changes in these miRNAs in response to the interventions were evaluated. The primary outcomes were changes in abdominal adipose tissue [visceral (VAT), deep subcutaneous (DSAT), and superficial subcutaneous (SSAT) adipose tissue; cm²] (n = 144). The secondary outcomes were changes in ectopic fat accumulation in the liver (n = 141) and pancreas (n = 143).

Results

Greater decreases in miR-100-5p were associated with more reductions of VAT (β ± SE per 1-SD decrease: −9.63 ± 3.13 cm²; P = 0.0025), DSAT (β ± SE: −5.48 ± 2.36 cm²; P = 0.0218), SSAT (β ± SE: −4.64 ± 1.68 cm²; P = 0.0067), and intrahepatic fat percentage (β ± SE: −1.54% ± 0.49%; P = 0.0023) after the interventions. Similarly, participants with greater decrease in miR-99a-5p had larger 18-mo reductions of VAT (β ± SE: −10.12 ± 3.31 cm² per 1-SD decrease; P = 0.0027) and intrahepatic fat percentage (β ± SE: −1.28% ± 0.52%; P = 0.015). Further, decreases in circulating miR-99b-5p (β ± SE: per 1-SD decrease: −0.44% ± 0.21%; P = 0.038) and miR-100-5p (β ± SE: −0.50% ± 0.23%; P = 0.033) were associated with a decrease in pancreatic fat percentage, as well as improved glucose metabolism and insulin secretion at 18 mo.

Conclusions

Decreases in circulating miR-99-5p/100-5p expression induced by lifestyle interventions were related to improved body fat distribution and ectopic fat accumulation. Our study suggests that changes in circulating adipose-related miR-99-5p/100-5p may be linked to reducing diabetogenic fat depots in patients with abdominal obesity.

This trial was registered at clinicaltrials.gov as NCT01530724.

Introduction

Excess visceral fat accumulation has been associated with insulin resistance and an elevated risk of type 2 diabetes (1, 2). Also, increased visceral fat reflects ectopic fat accumulation in other organs (such as the liver and pancreas) that are closely related to the pathogenesis of type 2 diabetes (2–5). MicroRNAs (miRNAs) are short noncoding RNAs that posttranscriptionally regulate gene expression (6–8). Circulating miRNAs have emerged as novel biomarkers of obesity and metabolic diseases that coordinate whole-body metabolism through intercellular communications (9–11). A study has shown that adipose tissue is a major source of circulating exosomal miRNAs, which can regulate gene expression in distant tissues and thereby serve as a previously undescribed form of adipokine (12).

MicroRNAs-99/100 (miR-99/100) are adipose tissue–related miRNAs (12–15) that have been suggested to play pivotal roles in controlling adiposity, weight gain, and liver steatosis, by regulating the expression of the adipogenesis-related peroxisome proliferator–activated receptor γ (PPARγ), adipose tissue inflammation, and other key factors of lipid storage and metabolism (14, 16, 17). The previous study showed that circulating miR-99b might regulate fibroblast growth factor 21 (FGF21) expression in the liver, suggesting adipose–hepatic crosstalk via circulating miR-99b (12). Other earlier studies have also reported that miR-99/100 are expressed in human adipose tissue, and these miRNAs are involved in adipogenesis and dysregulated in adipose tissue of subjects with obesity (13, 15, 18).

On the other hand, no clinical study has addressed how changes in the adipose-related circulating miR-99/100 induced by lifestyle interventions are associated with improved visceral adiposity and body fat distribution. Existing studies have also observed potential regulatory roles of miR-99/100 in liver fat metabolism and insulin/glucose metabolism (19–22). Nonetheless, associations of changes in circulating miR-99-5p/100-5p with long-term changes in diabetogenic ectopic fat depots (such as in the liver and pancreas), glucose metabolism, and insulin sensitivity in response to lifestyle interventions remain unknown.

Therefore, we investigated associations of changes in adipose-related circulating miR-99-5p/100-5p with the long-term reductions of visceral and ectopic fat depots over 18 mo among adults with abdominal obesity who participated in an 18-mo lifestyle intervention, the CENTRAL trial (23). We further examined associations of changes in these adipose-related circulating miRNAs with the improvements in glucose metabolism and insulin sensitivity in response to the lifestyle interventions over 18 mo.

Methods

Study participants

The CENTRAL trial (NCT01530724) is an 18-mo weight-loss lifestyle intervention trial that compared effects of lifestyle strategies on changes in specific body fat depots. The trial was conducted from October 2012 through April 2014 at the Nuclear Research Center Negev (Dimona, Israel), a workplace with a dedicated cafeteria and an on-site medical clinic. The study was approved and monitored by the human subjects committee of Soroka Medical Center. All participants provided written informed consent. Detailed information on the study design and methods has previously been described (23). In this trial, a total of 278 adults with abdominal obesity (75%) or dyslipidemia were randomly assigned to 1 of 2 equally hypocaloric diets [a low-fat (LF) diet or a Mediterranean/low-carbohydrate (MED/LC) diet] for the entire study period of 18 mo. The achievement goal of the LF diet was to reduce fat intake (30% of energy from fat, ≤10% from saturated fat, and ≤300 mg cholesterol/d) and to increase dietary fibers. The achievement goal of the MED/LC diet was restricting carbohydrate intake to <40 g/d during the first 2 mo and thereafter a gradual increase ≤70 g/d and increased protein and fat intake. The MED/LC diet was rich in vegetables and legumes and low in red meat, with poultry and fish replacing beef and lamb. This group was also provided 28 g walnuts/d starting from the third month. After 6 mo, each diet group was further randomly assigned into added physical activity (PA) groups (LF-PA+, MED/LC-PA+) or no added PA groups (LF-PA−, MED/LC-PA−) for the last 12 mo of the intervention. Exclusion criteria were the presence of impaired liver function, elevated serum creatinine, or active cancer; pregnancy or lactation; highly physically active or unable to take part in PA; or participation in another trial. The study investigators assessing the outcomes were blinded to the group assignments.

In the present analysis, 227 participants were eligible based on the availability of blood samples collected at baseline for the measurement of circulating miRNAs. As reported previously (23), 86% of the study participants completed the 18-mo trial with good adherence; a total of 157 participants had blood samples taken at 18 mo after the interventions for measurement of circulating miRNAs at 18 mo (Supplemental Figure 1). We confirmed that baseline expression of circulating miR-99/100 and baseline adiposity [BMI (in kg/m²) and visceral adipose tissue (VAT)] were not significantly different between participants with (n = 157) and those without (n = 70) data at 18 mo. Our study aimed to analyze associations of changes in miRNAs with changes in outcome traits over 18 mo; therefore, participants with missing data on exposures or study outcomes were excluded when performing each analysis. For the primary outcomes of changes in abdominal body fat distribution, a total of 144 participants were eligible for the analyses.

Measurements of body adiposity and fat distribution

Weight and height were measured without shoes; waist circumference was measured halfway between the last rib and the iliac crest to the nearest millimeter using an anthropometric measuring tape. Body fat distribution [VAT, deep subcutaneous adipose tissue (DSAT), superficial subcutaneous adipose tissue (SSAT)] and ectopic fat accumulation (liver, pancreas, pericardial, renal, and muscle) were assessed by whole-body MRI at baseline and 18 mo. Total adipose tissue (TAT) was calculated as the sum of VAT, DSAT, and SSAT. MRI was performed using a 3-Tesla magnet (Ingenia 3.0 T, Philips Healthcare); details on the MRI and the imaging analysis have been described elsewhere (23). Intrahepatic fat percentage was assessed as the percentage of liver fat as a whole, using a standard method described previously (24). Pancreatic fat percentage was quantified by a semiautomatic MATLAB-based program, analyzing the average of 3 successive 2-dimensional slices, each including all pancreatic regions (23, 25). Renal (left + right) sinus fat area (cm²) and fat accumulation in renal parenchyma (%) were also measured using the program (23, 26); both parameters of fat storage in the kidney were related to abdominal adiposity in this trial (26). Intermuscular fat (i.e., adipose tissue located within and between the thigh muscles and separated from subcutaneous adipose tissues) was determined using a single axial 2D fat-phase image (2 mm thickness) of the right midthigh (27). Pericardial fat accumulation was evaluated in 29% (n = 80 of 278) of the original study participants, double-blindly randomly selected (28); changes in extrapericardial fat (i.e., pericardial fat that lies external to the fibrous pericardium) were analyzed in a subcohort (n = 38) of the present study's participants.

Measurements of biomarkers

Fasting blood samples were collected at baseline and after 18 mo of the intervention, and stored at −80°C. Biomarker measurements were performed at the Leipzig University laboratories, Germany. Fasting glucose was measured by Roche GLUC 3 (hexokinase method); fasting insulin was measured with the use of an enzyme immunometric assay (Immulite automated analyzer, Diagnostic Products). HOMA-IR and HOMA of β-cell function were calculated. We also measured circulating liver-fat-related biomarkers [retinol-binding protein 4 (RBP4) and ferritin] at baseline and 18 mo. RBP4 is secreted by hepatocytes and adipose tissue (29), and serum ferritin concentrations are commonly elevated in patients with nonalcoholic fatty liver disease owing to systemic inflammation, increased iron stores, or both conditions (30). Changes in TAT or VAT were positively correlated to changes in fasting insulin, HOMA-IR/HOMA of β-cell function, and ferritin concentrations (with Pearson correlation coefficients of 0.20–0.31; P < 0.01 for all) (Supplemental Table 1); there were no significant correlations between changes in TAT or VAT and changes in fasting glucose or RBP4 concentrations. Detailed associations of VAT and metabolic traits in the original study participants have been described previously (23).

Assessment of miRNAs

We extracted total serum RNA using the RNeasy Serum/Plasma Advanced kit (Qiagen). We used ≤20 ng of serum total RNA in the small RNA protocol of the NEBNext® Small RNA Library Prep Set for Illumina according to the manufacturer's instructions (NEB). The barcoded libraries between 140 and 165 bp were purified and quantified using the Library Quantification Kit-Illumina/Universal (KAPA Biosystems). In addition, the size distribution of the miRNA libraries was visualized on a Fragment Analyzer (Agilent). A total of 227 samples at baseline and 157 samples at 18 mo after the intervention were eligible with sufficient quality to proceed to next-generation sequencing. A pool of ≤100 libraries was used for cluster generation at a concentration of 1.5 pM followed by sequencing of 75 bp using an Illumina NextSeq 550 sequencer at the sequencing core facility of Leipzig University (Faculty of Medicine) using version 2.5 flowcell and chemistry according to the manufacturer's instructions (Illumina). Demultiplexing of raw reads, adapter trimming, and quality filtering were conducted using Illumina bcl2fastq conversion (version 2.20.0) and cutadapt software (version 1.18). Mapping against the human reference genome (hg38) and miRbase reference sequences (version 22) was conducted using Bowtie2. Read counts were calculated with the Rsamtools (version 2.0.1-1) R bioconductor package; normalized data were used for the analysis. In this study, we analyzed circulating has-miR-100-5p, has-miR-99a-5p, and has-miR-99b-5p at baseline and 18 mo. miR-100 belongs to the miR-99 family of miRNAs (i.e., human miRNA family miR-99-5p/100-5p with seed ACCCGUA). A total of 156 participants had data on miR-100-5p at both baseline and 18 mo; 157 participants had data on miR-99a-5p/99b-5p at the 2 time points in this study.

Statistical analysis

Data on miR-99/100 were log-transformed to improve data normality before calculating changes. We first analyzed relations between miR-99/100 expression and adiposity and other characteristics at baseline. The primary exposures were changes in circulating miR-99/100 expression over 18 mo. To better present the magnitude of the miRNA changes in response to the intervention, we used the percentage change (i.e., fold change) of each miRNA based on the ratio of 18-mo change after the intervention to the baseline value (Supplemental Figure 2). The primary outcomes were 18-mo changes in body fat distribution (particularly VAT as the main primary outcome) after the dietary intervention. We tested interactions between miRNA changes and diet or PA intervention groups for the primary outcomes to check whether the associations were significantly different across the dietary or PA intervention groups. Secondary outcomes were changes in ectopic fat depots in the liver and pancreas (i.e., intrahepatic fat percentage and pancreatic fat percentage). To confirm the findings on the primary and secondary outcomes, we analyzed other adiposity measures (such as weight, waist circumference, renal parenchyma fat, renal sinus fat, intermuscular fat, and extrapericardial fat). In addition, we also examined roles of the adipose-related miR-99/100 in relation to changes in liver-fat-related biomarkers (RBP4 and ferritin), glucose metabolism, and insulin sensitivity. General linear models were performed to calculate β ± SE for the respective outcomes per 1-SD decrease in miR-99/100 change (% change from baseline to 18 mo) after the interventions after adjusting for age, sex, intervention group, the individual miRNA expression value at baseline, and the respective outcome trait at baseline. Sensitivity analyses were performed further adjusting for baseline BMI or waist circumference in the model. We also performed a sensitivity analysis using models further adjusting for baseline and changes in alcohol consumption. Trajectories of changes in body fat distribution over 18 mo were analyzed according to the tertile (T) categories of miR-99a or miR-100; the lowest tertile (T1) category included participants with the largest reduction in the respective miRNA. We used linear mixed models including time × tertiles of miR-100-5p or miR-99a-5p to test the miRNA effect on the trajectory of changes in VAT and intrahepatic fat. Statistical analyses were performed with SAS version 9.3 (SAS Institute). P < 0.05 was considered a statistically significant level.

Results

Table 1 shows characteristics of the study participants according to baseline expression of miR-100. At baseline, participants with higher miR-100 expression had greater VAT (P-trend = 0.08) and intrahepatic fat accumulation (P-trend < 0.0001). Also, higher expression levels of miR-99a (P-trend < 0.0001) were related to greater degrees of intrahepatic fat (P-trend < 0.0001) and HOMA-IR (P-trend = 0.04) at baseline (Supplemental Table 2).

TABLE 1.

Characteristics of the study participants according to tertile categories of circulating miR-100-5p expression at baseline¹

	miR-100-5p
Variables	T1 (n = 75)	T2 (n = 76)	T3 (n = 75)	P*
Age, y	47.1 ± 9.7	47.0 ± 9.4	49.6 ± 9.1	0.10
Male	67 (89.3)	66 (86.8)	66 (88.0)	0.89
BMI, kg/m2	30.9 ± 3.9	30.4 ± 3.6	31 ± 3.8	0.92
Weight, kg	91.5 ± 13.3	90.8 ± 13.5	90.5 ± 12.0	0.63
Waist circumference, cm	107 ± 10	106 ± 9	107 ± 9	0.88
Abdominal adipose tissue area, cm2
Visceral adipose tissue	163.5 ± 63.7	175.4 ± 74.3	182.8 ± 61.4	0.08
Deep subcutaneous adipose tissue	211.4 ± 67.6	211.3 ± 67.2	221.8 ± 71.5	0.36
Superficial subcutaneous adipose tissue	140.6 ± 56.7	138.1 ± 52.5	146.1 ± 70.4	0.58
Total adipose tissue	515.5 ± 136.8	524.8 ± 149.5	550.7 ± 147.4	0.14
Intrahepatic fat, %	7.2 ± 8.6	8.8 ± 9.2	13.9 ± 11.4	<0.0001
Pancreatic fat, %	17.2 ± 5.3	16.7 ± 4.9	17.2 ± 4.6	0.94
Renal parenchyma fat, %	7.7 ± 1.6	7.8 ± 1.8	8.1 ± 2.2	0.18
Renal sinus fat, cm2	2.5 ± 1.3	2.6 ± 1.3	2.9 ± 1.5	0.06
Intermuscular fat, cm2	9.6 ± 4.8	8.8 ± 4.5	10.5 ± 4.9	0.25
Extrapericardial fat, mL	191.7 ± 71.7	200.9 ± 86.6	195.9 ± 73.4	0.85
Fasting glucose, mg/dL	106.8 ± 18.1	107.6 ± 16.7	107.5 ± 23.4	0.85
Fasting insulin, μIU/mL	17.1 ± 11.7	17.3 ± 11.0	16.9 ± 9.2	0.89
Log-transformed HOMA-IR	1.3 ± 0.6	1.4 ± 0.5	1.3 ± 0.6	0.94
Log-transformed HOMA of β-cell function	4.8 ± 0.5	4.9 ± 0.6	4.8 ± 0.6	0.98
Retinol-binding protein 4, μg/mL	50.6 ± 31.8	44.8 ± 26.3	45.0 ± 30.3	0.33

¹Values are mean ± SD or n (%) unless otherwise indicated. miR-100, microRNA-100; T, tertile.

*P value by chi-square test for sex or P-trend across the tertile categories for continuous variables by general linear model.

In response to the 18-mo lifestyle intervention, changes in miR-100 (range of change from baseline to 18 mo: −21% to 29%), miR-99a (range: −14% to 29%), and miR-99b (range: −19% to 15%) showed considerable interindividual variability across the study participants (Supplemental Figure 2). The degrees of miRNA changes did not differ by the types of interventions (such as MED/LC diet compared with LF diet, or PA+ compared with PA− intervention) (Supplemental Table 3). We then tested associations of changes in miR-99/100 after the interventions with changes in body fat distribution and overall adiposity over 18 mo (Table 2). We found that greater decreases in miR-100 were significantly associated with decreases in fat depots (VAT, P = 0.0025; DSAT, P = 0.0218; SSAT, P = 0.0067), as well as reductions of TAT (P = 0.0019) and weight (P = 0.0028). For example, each 1-SD decrease of miR-100 (which was equivalent to per −9.18% from the initial value) was associated with a VAT reduction of −9.63 cm² (SE: 3.13 cm²) at 18 mo. Similarly, each 1-SD decrease of miR-99a (which was equivalent to per −6.29% from the initial value) was associated with reductions of VAT (β ± SE: −10.12 ± 3.31 cm²; P = 0.0027), DSAT (β ± SE: −7.01 ± 2.50 cm²; P = 0.0057), SSAT (β ± SE: −4.45 ± 1.79 cm²; P = 0.014), and TAT (β ± SE: −21.62 ± 6.53 cm²; P = 0.0012). There were no significant interactions between miR-99/100 changes and different diet groups or PA intervention groups (P-interaction > 0.05 for all) for these outcomes.

TABLE 2.

Associations of changes in circulating miR-99/100 after a diet/lifestyle intervention with changes in body fat distribution and overall adiposity over 18 mo¹

	miR-100-5p change			miR-99a-5p change			miR-99b-5p change
Outcomes (changes from baseline to 18 mo)	n	β ± SE	P	n	β ± SE	P	n	β ± SE	P
Adipose tissue area, cm2
Visceral adipose tissue	144	−9.63 ± 3.13	0.0025	144	−10.12 ± 3.31	0.0027	144	−4.85 ± 2.93	0.1
Deep subcutaneous adipose tissue	144	−5.48 ± 2.36	0.0218	144	−7.01 ± 2.50	0.0057	144	−0.29 ± 2.24	0.9
Superficial subcutaneous adipose tissue	144	−4.64 ± 1.68	0.0067	144	−4.45 ± 1.79	0.014	144	−2.61 ± 1.58	0.1
Total adipose tissue	144	−19.6 ± 6.19	0.0019	144	−21.62 ± 6.53	0.0012	144	−7.48 ± 5.88	0.21
Weight, kg	150	−1.55 ± 0.51	0.0028	150	−0.96 ± 0.53	0.07	150	−0.47 ± 0.47	0.32
Waist circumference, cm	141	−1.09 ± 0.58	0.06	141	−0.86 ± 0.61	0.16	141	−0.42 ± 0.53	0.43

¹General linear models were performed to calculate β ± SE per 1-SD decrease in each miRNA change (% change from baseline to 18 mo after the intervention) for the respective outcomes after adjusting for age, sex, intervention group, the respective miRNA at baseline, and the respective outcome trait at baseline. Definition of 1-SD reduction of change after the intervention: −9.18% for miR-100-5p; −6.29% for miR-99a-5p; −5.97% for miR-99b-5p. miRNA, microRNA; miR-99/100, microRNA-99/100.

When we investigated associations of miR-99/100 with ectopic fat accumulation in the liver and pancreas (Figure 1), reductions of intrahepatic fat depots at 18 mo were related to greater decreases in miR-100 (β ± SE: −1.54% ± 0.49% per 1-SD decrease; P = 0.0023) or miR-99a (β ± SE: −1.28% ± 0.52% per 1-SD decrease; P = 0.015). For changes in pancreatic fat percentage, greater decreases in miR-100 (β ± SE: −0.50% ± 0.23% per 1-SD decrease; P = 0.033) and miR-99b (β ± SE: −0.44% ± 0.21%; P = 0.038) showed significant associations with decreases in pancreatic fat percentage.

FIGURE 1.

Associations of changes in circulating miR-99/100 after a diet/lifestyle intervention with 18-mo changes in intrahepatic fat (A) and pancreatic fat (B). Data are β ± SE per 1-SD decrease in each miRNA change (% change from baseline to 18 mo after the intervention) for changes in (A) intrahepatic fat (n = 141) or (B) pancreatic fat (n = 143) after adjusting for age, sex, intervention group, the respective miRNA at baseline, and the respective outcome trait at baseline using general linear models. The definition of 1 SD decrease: −9.18% for miR-100-5p; −6.29% for miR-99a-5p; −5.97% for miR-99b-5p. miRNA, microRNA; miR-99/100, microRNA-99/100.

Because we found that changes in miR-100 or miR-99a were significantly associated with improved VAT and intrahepatic fat at the end of the intervention (18 mo), we also examined trajectories of changes in these outcomes according to the tertile categories of miR-100-5p or miR-99a-5p changes (Figure 2). Participants who showed the largest reduction of miR-100 [in the lowest tertile (T1) group: median (IQR): −5.0% (−9.8%, −2.7%)] or miR-99a [in the T1 group: median (IQR): −4.6% (−6.8%, −2.7%)] had improved VAT (P_{time × miR-100-interaction} = 0.033 in Figure 2A; P_{time × miR-99a-interaction} = 0.014 in Figure 2C) and intrahepatic fat (P_{time × miR-100-interaction} < 0.001 in Figure 2B; P_{time × miR-99a-interaction} = 0.005 in Figure 2D) over 18 mo; the higher 2 tertile groups (T2 and T3) had consistently lesser reductions of these outcomes over 18 mo than the T1 group.

FIGURE 2.

Trajectories of changes in VAT and intrahepatic fat over 18 mo according to the tertile categories of miR-100-5p changes (A, B) or miR-99a-5p changes (C, D). Total number of participants: n = 87 at 6 mo; n = 144 at 18 mo. For the tertiles of miR-100-5p changes (% change from baseline to 18 mo after the intervention), median (IQR) values were as follows: T1 (dashed): −5.0% (−9.8%, −2.7%); T2 (gray): 2.5% (1.4%, 5.1%); and T3 (black): 12.4% (8.1%, 16.4%). For the tertiles of changes in miR-99a-5p (% change from baseline to 18 mo after the intervention), median (IQR) values were as follows: T1 (dashed): −4.6% (−6.8%, −2.7%); T2 (gray): 1.6% (0.4%, 2.6%); and T3 (black): 7.1% (5.4%, 10.4%). P_{time × miRNA} interactions by the linear mixed models: (A) P_{time × miR-100-interaction} = 0.033; (B) P_{time × miR-100-interaction} < 0.001; (C) P_{time × miR-99a-interaction} = 0.014; (D) P_{time × miR-99a-interaction} = 0.005. miR-99/100, microRNA-99/100; T, tertile; VAT, visceral adipose tissue.

We further performed several sensitivity analyses. In models controlling for baseline BMI or waist circumference (Supplemental Table 4), there were similarly significant relations between changes in the respective miRNAs and changes in body fat distribution and ectopic fat. In the present study, almost all of the study participants did not report excessive alcohol consumption, and we confirmed that the associations of miR-99/100 with liver/pancreatic fat were not altered when adjusting for baseline and changes in alcohol consumption (Supplemental Figure 3). When we performed additional analyses on changes in ectopic fat depots in other organs (Supplemental Figure 4), we found that decreases in miR-100 were consistently and significantly associated with decreases in renal parenchyma fat (P = 0.0004), renal sinus fat (P = 0.0027), intermuscular fat (P = 0.0018), and extrapericardial fat (P = 0.049) at 18 mo.

Finally, we analyzed whether changes in the adipose-related miR-100 and miR-99b were related to changes in glucose metabolism, insulin sensitivity, and liver biomarkers (RBP4 and ferritin) (Table 3). We found that decreases in miR-100 were related to reductions in fasting glucose (P = 0.031), RBP4 (P = 0.02), and ferritin (P = 0.009) concentrations at 18 mo. Greater decreases in miR-99b were significantly associated with larger increases in HOMA of β-cell function (P = 0.004) and fasting insulin (P = 0.026) at 18 mo. Participants with greater decreases in miR-99b were characterized as having lower pancreatic fat, fasting glucose, and higher HOMA of β-cell function at the end of the intervention (Supplemental Figure 5). In sensitivity analyses, associations of miR-100 with fasting glucose and RBP4 were independent of concurrent changes in TAT or VAT. Also, the relation between changes in miR-99b and HOMA of β-cell function was independent of concurrent changes in TAT or VAT (Supplemental Table 5) when these variables (i.e., changes in miR-99b and changes in VAT or TAT) were concurrently included in the models. Changes in VAT (P = 0.028) or TAT (P = 0.007) were also significantly related to HOMA of β-cell function changes, independently of miR-99b changes in the concurrently adjusted models.

TABLE 3.

Associations of changes in miR-100 and miR-99b with 18-mo changes in liver-fat-related biomarkers, glucose metabolism, and insulin sensitivity¹

	miR-100-5p change			miR-99b-5p change
Outcomes (changes from baseline to 18 mo)	n	β ± SE	P	n	β ± SE	P
Fasting glucose	156	−2.81 ± 1.29	0.031	157	−1.28 ± 1.14	0.26
Log-transformed fasting insulin	155	−0.03 ± 0.04	0.38	156	0.08 ± 0.03	0.026
Log-transformed HOMA-IR	155	−0.06 ± 0.04	0.19	156	0.06 ± 0.04	0.085
Log-transformed HOMA of β-cell function	155	0.02 ± 0.04	0.57	156	0.1 ± 0.03	0.004
Retinol-binding protein 4	148	−0.14 ± 0.06	0.02	149	−0.02 ± 0.05	0.76
Ferritin	154	−0.13 ± 0.05	0.009	155	−0.06 ± 0.04	0.14

¹General linear models were performed to calculate β ± SE per 1-SD decrease in each miRNA change (% change from baseline to 18 mo after the intervention) for the respective outcome after adjusting for age, sex, intervention group, the respective miRNA at baseline, and the respective outcome trait at baseline. Definition of 1-SD reduction of change after the intervention: −9.18% for miR-100-5p; −5.97% for miR-99b-5p. miRNA, microRNA; miR-99/100, microRNA-99/100.

Discussion

The present study found that decreases in circulating adipose-related miR-99a and miR-100 induced by diet and lifestyle modifications were associated with reduced visceral adiposity and intrahepatic fat accumulation over 18 mo. In addition, decreases in circulating miR-99b and miR-100 were associated with greater reductions in pancreatic fat and more evident improvements in parameters of glucose metabolism and insulin sensitivity. Our findings suggest that changes in circulating adipose-related miR-99/100 may be linked to reductions of diabetogenic fat depots and metabolic disorders during lifestyle interventions among adults with abdominal obesity.

Our study is the first that we know of to show that long-term changes in circulating adipose-related miR-99-5p/100-5p induced by diet/lifestyle interventions were associated with decreases in various body fat and ectopic fat depots in adults with abdominal obesity. Recent evidence demonstrates that a significant proportion of circulating miRNAs originate from adipose tissues (9, 12). Adipose tissue–specific knockout of the miRNA-processing enzyme Dicer (ADicerKO) mice have a defect in miRNA processing specifically in adipose tissue and are also characterized as having decreased white adipose tissue mass (31) and reductions of circulating exosomal miRNAs (including miR-99a/99b/100) (12). Our study participants tended to reduce VAT (the primary outcome) over 18 mo after the interventions (23); it would be therefore reasonable to conclude that the reductions of adipose-related miR-100 and miR-99a were strongly related to decreases in abdominal adipose tissue, in particular VAT. Although it was challenging to infer the causality of the associations, our findings are also supported by prior studies showing that miR-99/100 may regulate adipose tissue, adipose-related inflammation, and lipid storage and metabolism (12–17). In addition, we found that circulating miR-99/100 changes in response to the interventions were associated with reductions of diabetogenic ectopic fat depots, such as in the liver and pancreas. A previous study in diabetic mice suggests a potential role of miR-99a in pathological changes in liver lipid metabolism (19), and a cross-sectional study also suggests that miR-99/100 expression in VAT (in particular, miR-99b) may be linked to the pathogenesis of fatty liver disease (32). We found that participants with decreases in miR100 and miR-99a in response to the intervention had consistently improved intrahepatic fat content over 18 mo. In addition, fatty pancreas, an accumulation of intrapancreatic fat, is frequently observed in overweight and obese individuals with impaired glucose metabolism (33–35). We found that the miR-99b expression was significantly related to reducing fat in the pancreas, although miR-99b did not show associations with changes in general obesity, which is in part consistent with studies suggesting that circulating miR-99b-5p was not a marker of weight loss after bariatric surgery (36, 37). Of note, changes in circulating miR-100 showed significant associations with improved ectopic fat depots in various organs or tissues (such as hepatic, renal, intermuscular, and extrapericardial fat depots). These findings on the ectopic fat changes are supported by the previous study suggesting that the ADicerKO mice had different expression of Fgf21 mRNA in liver, muscle, and pancreas (12) as compared with wild-type mice. FGF21 has been implicated as a whole-metabolic regulator (38, 39) and FGF21 may be a target of miR-99/100 (12, 40).

Further, our findings that miR-99/100 changes were related to changes in liver biomarkers (RBP4 and ferritin) and glucose/insulin metabolism lent further support to the role of these miRNAs as circulatory adipokines. Existing evidence indicates that hepatic steatosis alters the secretion of hepatokines including RBP4 to promote insulin resistance and other metabolic complications (5). In addition, it has been reported that ADicerKO mice were characterized as having insulin resistance and elevated concentrations of branched-chain amino acids (12, 41). Some studies have also reported that insulin-like growth factor 1 receptor (IGF1R)/insulin growth factor receptor (IGFR) may be a target of miR-100 (18, 42) and that miR-99b/100 expression in plasma or adipose tissue may be associated with type 2 diabetes or prediabetes (20–22). In the present study, the relations between miR-100 and fasting glucose or RBP4 changes were not affected by concurrent VAT or TAT changes. Also, the relation between miR-99b and HOMA of β-cell function was independent of visceral or total adiposity changes, even though the changes in adiposity were correlated to changes in HOMA of β-cell function in this population. These results suggested that the concurrent changes in overall adiposity did not fully explain the associations; the robust relation between miR-99b and HOMA of β-cell function was in line with the significant results on changes in miR-99b and ectopic fat (in particular pancreatic fat). Further studies are warranted to explore novel biological pathways underlying the findings and to investigate whether the specific adipose-related miRNAs could be therapeutic targets in regulating liver/pancreatic steatosis, secretion of other adipokines, and related metabolic dysregulation.

There are several strengths in the present study. We assessed changes in circulating miRNAs with longitudinal repeated measurements in a long-term lifestyle intervention trial with an adequate sample size, which was thus far the largest of its kind. In particular, the primary endpoint of the CENTRAL trial was visceral adiposity, and the study design was ideal for testing the present study hypothesis. The robust and consistent results on the miR-99/100 family and various ectopic fat depots measured by whole-body MRI scans would strengthen our conclusion. Several limitations should also be considered. Our study mainly included male participants with abdominal obesity, and whether the findings could be applicable to cohorts of women or the general population needs to be further confirmed in other studies. Also, it was challenging to infer the causality of the associations because the changes in circulating miRNAs and changes in outcome measurements were assessed concurrently. Our study did not assess the timing and trajectories of miRNA changes contributing to the body fat depot reductions. To establish whether changes in circulating miRNAs may be predictive of the subsequent outcome changes would require further investigation. Also, further experimental investigations are warranted to verify the causality of the relations observed in this study.

In conclusion, this study showed that changes in circulating miR-99/100 expression in response to diet/lifestyle interventions were significantly associated with decreases in visceral/ectopic fat depots, as well as glucose metabolism, among patients with abdominal obesity. The specific adipose-related circulating miRNA changes may be potential targets linked to improving visceral and ectopic fat accumulation and metabolic health during lifestyle interventions.

Notes

Supported by NIH National Heart, Lung, and Blood Institute grants HL071981, HL034594, and HL126024 (to LQ); NIH National Institute of Diabetes and Digestive and Kidney Diseases grants DK091718, DK100383, and DK115679 (to LQ); Tulane Research Centers of Excellence Awards (to LQ); and US-Israel Binational Science Foundation grant 2011036. LQ was supported by American Heart Association Scientist Development Award 0730094N. YH was supported by a Japan Society for the Promotion of Science Grant-in-Aid for Scientific Research 14J30007 and Postdoctoral Fellowship for Research Abroad and the 2019 American Heart Association postdoctoral fellowship award 19POST34380035. The CENTRAL study was supported by Deutsche Forschungsgemeinschaft (DFG; German Research Foundation) grants 209933838—SFB 1052: B01 (to MB), B03 (to PK), B11 (to IS), Z04 (to IS and UC) and by Deutsches Zentrum für Diabetesforschung (DZD) grant 82DZD00601, the Israel Science Foundation, Israel Ministry of Science and Technology grant 3-13604 (to IS), and the Dr Robert C and Veronica Atkins Research Foundation. The sponsors had no role in the design or conduct of the study.

Supplemental Figures 1–5 and Supplemental Tables 1–5 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/ajcn/.

Abbreviations used: ADicerKO, adipose tissue–specific Dicer knockout; DSAT, deep subcutaneous adipose tissue; FGF21, fibroblast growth factor 21; LF, low-fat; MED/LC, Mediterranean/low-carbohydrate; miRNA, microRNA; miR-99/100, microRNA-99/100; PA, physical activity; RBP4, retinol-binding protein 4; SSAT, superficial subcutaneous adipose tissue; T, tertile; TAT, total adipose tissue; VAT, visceral adipose tissue.

Data Availability

Data described in the article, code book, and analytic code will be available from the corresponding author on reasonable request. All data generated or analyzed during this study are included in this article (and its supplementary information files).